S. C. Colbeck

Snow Particle Morphology U.S. Army Cold Regions Research and Engineering Laboratory in the Seasonal Cover Hanover, N.H. 03755

Abstract Perhaps the striking hexagonal symmetry, the delicate frame, the high visibility of falling snow, and the multitude of pho- Snow precipitation degenerates rapidly once it reaches the ground. tographs (e.g., Bentley and Humphreys, 1931) account for A wide variety of particle types develop in seasonal snow covers, the general awareness of the morphology of snowflakes. thus leading to a wide range of snow properties. The most common Systematic observations and explanations of seasonal snow varieties of particles are shown here. The physical processes respon- sible for the growth and development of these particles are described morphology began in the Swiss Alps with the work of Selig- in general terms, although these processes are not understood as well man (1936) and Bader et al (1939). Interest grew interna- as the processes of growth in the atmosphere. The heat and tionally (e.g., de Quervain, 1954; Yosida et al. 1955; Kuzmin, mass flows associated with the development of these in the 1957; Giddings and LaChapelle, 1962) because of the aware- snow cover are complicated because of snow's complex geometry. ness of the great variations in snow properties with changes in particle shape. Much of this earlier work was centered on the transition from the highly rounded to the highly faceted crys- tals shown in Figs. 2 and 3, respectively. Many other snow 1. Introduction particle types are equally interesting (see Colbeck, 1982), al- though work on them has been even more recent. The most An introduction to the wide variety of particle morpholo- important forms are outlined in Table 1 and described later; gies in the seasonal snow cover is given here. Because a wide some suggestions are given about their formation. While it is range of morphologies normally develops, the important clear that interest in their growth has lagged interest in the material properties of the seasonal snow cover also are highly growth of snowflakes and hailstones in the atmosphere, our variable. That the seasonal snow undergoes very active and understanding of the particle morphology of snow cover is continuous changes should not be surprising; the snow is often making rapid advances through use of information generated at or close to its melting point, the particles are in intimate by cloud physicists (e.g., Hobbs, 1974) and crystallographers. contact, the surface area per unit volume is very large, and The outline of seasonal snow morphology given in Table 1 the snow surface experiences constantly changing thermal is, like previous attempts to categorize snow, arbitrary and conditions. This is a sure prescription for constant change or subject to revision as we increase our understanding of the metamorphism. In fact, snow usually begins "destructive processes and conditions. It begins logically with the source, metamorphism" as soon as the precipitation reaches the but then divides snow on the ground arbitrarily into three ground. Figure 1 shows a broken snowflake in an early stage categories—dry snow, wet snow, and a menagerie of mixed of the process in which it loses its sharp crystalline faces and processes. Most snow particles can be classified adequately develops well-rounded branches. Eventually this process as Ha, lib, Ilia, or Illb, although the other groups can be leads to the well-rounded crystals shown in Fig. 2. quite important for particular purposes. The headings of the Perhaps the best known and most dramatic metamorphism subsections of Sections 3,4, and 5 correspond with the lettered is the recrystallization from well-rounded (Fig. 2) to highly forms listed under II, III, and IV, respectively, in Table 1. faceted crystals (Fig. 3). These and other ice particle types are described later. First, a brief history of observations of the particles in the snow cover is given.

3. Dry snow

a. Equilibrium form 2. Snow cover observations At slow growth rates, crystal forms develop that minimize The relatively recent notice of the rich variety of particle the total surface free energy of an entire crystal (see the Ap- types in the seasonal snow cover (e.g., Wolley, 1858) is in pendix—Glossary of terms). At these slow rates, crystals sharp contrast to over 400 years of observations of snow grow by vapor condensation on the surface followed by the crystals grown in the atmosphere (see Frank, 1982). Not only incorporation of molecules into vacant sites on the surface, a have the observations lagged, but our understanding of the process similar to the growth of liquid droplets. As we might physical processes has lagged as well. For example, the well- expect, the shape of a crystal formed at very slow growth known Nakaya (1954) diagram displays the different crystal rates can be described approximately by equilibrium ther- forms of snow crystals as dictated by the temperature and modynamics, and the shape of such a crystal can be deter- supersaturation in the atmosphere, but reveals nothing about mined through the construction of a Wulff plot (see Appen- the growth of the highly rounded crystals shown in Fig. 2. dix). The equilibrium shape of an ice crystal in air generally is

602 Vol. 64. No. 6. June 1983

Unauthenticated | Downloaded 10/03/21 11:38 PM UTC Bulletin American Meteorological Society 603

TABLE 1. Outline of processes and conditions determining seasonal snow morphology.

I. Snow precipitate II. Dry snow a) Equilibrium form b) Kinetic growth form c) Transitional forms III. Wet snow a) Low liquid content b) High liquid content IV. Mixed processes a) Melt-freeze grains b) Surface melt crystals c) Wind and sun crusts d) Melt-freeze layers

FIG. 1. A broken snowflake in an early stage of "destructive met- amorphism.,, The sharp crystalline faces are disappearing as the sur- faces become rounded. thought to be a thick hexagonal plate, but, at higher tempera- tures, the equilibrium form is well rounded. Apparently, the surface transition layer, or Faraday's "liquid-like layer," re- duces the surface energy differences among the crystallo- graphic faces. The exact shape of the equilibrium form is not well established, although the slowly growing crystal in Fig. 2 suggests a prolate spheroid. This elongated shape commonly is observed in snow and may be the true equilibrium shape at temperatures just below the melting point. Sintering, or the movement of molecules to the interparticle contacts, occurs readily during the growth of the well-rounded form and gives snow its strength. As described next, sintering is reduced no- ticeably at large growth rates, thus leading to the cohesionless depth hoar crystals shown in Fig. 3.

b. Kinetic growth form

The growth rate of in a dry snow cover decreases with increasing snow density, but increases with the tempera- ture gradient through the snow cover. Apparently the super- saturation over the bottom side of an ice crystal is determined largely by the temperature difference between it and the crys- FIG. 2. Highly rounded, equilibrium form of snow crystals. This tals) just below it. Thus, the vapor pressure or supersatura- form dominates in seasonal snow at high temperatures and low tion increases with the temperature difference between indi- temperature gradients. These grains are bonded together by sintering. vidual crystals, which, in turn, increases with crystal spacing and temperature gradient. The crystal spacing is greater for less dense and the temperature gradient is greater for shallow snow covers during periods of cold weather (the ground surface normally is warmer than the snow surface during winter). If the supersaturation over the growing crystal exceeds a critical value, the growth proceeds rapidly by the propagation of distinct layers across the crystalline faces. The rate of movement of these layers varies with orientation in the crystal, so that some faces tend to grow out more quickly than others. This gives rise to the highly faceted crystals shown in Fig. 3, a clear departure from the well-rounded form shown in Fig. 2. These rapidly growing crystals are called the kinetic growth form because the surface kinetics determines their shape. Since equilibrium thermodynamics has little to do with their growth, they do not readily develop interparticle strength by FIG. 3. Highly faceted, kinetic growth form of snow crystals. sintering. Accordingly, a layer of these highly faceted crystals This form dominates at high temperature gradients. These depth (usually called depth hoar) can destabilize a snow cover and hoar crystals are cohesionless because of the lack of sintering during rapid growth. lead to . These faceted crystals grow most rapidly

Unauthenticated | Downloaded 10/03/21 11:38 PM UTC 604 Vol. 64, No. 6, June 1983

FIG. 5. Surface hoar crystals. These grow on cold, clear nights by vapor flow downward to the snow surface.

c. Transitional form

Things seldom fall neatly into two classes in nature, and it is FIG. 4. Chain of grains. At larger temperature gradients, the not surprising that crystals with mixed facets and rounded snow cover recrystallizes, with faceted crystals replacing rounded portions can be observed. Apparently some crystallographic crystals. This process starts with the growth of facets on the lower- most, coldest portions of chains of crystals, as indicated by arrows. faces develop facets more readily than others and produce mixed crystals at intermediate growth rates (see Fig. 6). These intermediate growth rates probably correspond to in the lower layers because both the diffusion of vapor temperature gradients that are close to the critical value men- in air and the vapor pressure difference between sources and tioned previously. sinks increase with mean temperature. During the most ac- Another version of the mixed crystal (Fig. 7) appears when tive growth, the highly faceted crystals are as cohesionless as rapid growth ceases. After developing as a fully faceted crystal oven-dried sand, while at intermediate growth rates they can at a high growth rate, a depth hoar crystal can lose its sharp retain some strength. corners and edges to rounding, much as the snowflake shown The actual mechanism for crystal growth is vapor transfer in Fig. 1. The onset of warmer weather reduces the tempera- among neighboring ice grains due to temperature differences, or Yosida et aVs (1955) "hand-to-hand delivery of water vapor" upward through the snow. At lower temperature gradients the vapor transfer is slower, so that the growth rate is slower and the well-rounded equilibrium form dominates. When the temperature gradient exceeds 10-20°C/m (de- pending on snow conditions), the rounded grains disappear and are replaced by faceted grains without a noticeable in- crease in snow density. This process begins with the growth of faceted crystals on the lowermost, coldest portions of chains of grains, such as the one shown in Fig. 4. Another type of highly faceted crystal (Fig. 5) grows on the snow surface and is called "surface hoar." These crystals form on cold, clear nights when the snow surface is cooled by radiating into space. Vapor moves downward to the cold sur- face from the warmer air and rapid crystal growth occurs. These crystals also are poorly bonded and, once buried by subsequent snowfalls, can become surfaces for future ava- FIG. 6. Crystals with mixed faceted and rounded portions at an lanche release. intermediate growth rate.

Unauthenticated | Downloaded 10/03/21 11:38 PM UTC Bulletin American Meteorological Society 605

next section, melt-freeze cycles are important, but clusters form and grains grow without these cycles (see Fig. 8). Grain growth by cannibalism, the consumption of smaller grains by larger grains, occurs in both wet and dry snows, but for very different reasons. In wet snow, the snow mass is macroscopically isothermal and the temperature variations occur only locally among the grains, with the smaller particles being at a lower temperature. The local temperature differ- ences may be small, but the path length for heat flow among neighboring grains is also small, so the heat flow can be large. Thus, the small grains melt rapidly as the large grains grow, and, as the smaller sources disappear, grain growth quickly slows. This typically happens at a grain size of 0.5-1 mm. In dry snow, the vapor pressure at thermodynamic equilib- rium over the ice crystals decreases with increasing size, whereas in wet snow, the melting temperature is lower for smaller particles. Since dry snow is always subjected to FIG. 7. Decaying depth hoar crystal. This crystal grew as a highly changing surface conditions and therefore to an ever-chang- faceted depth hoar crystal before the onset of warmer weather. Once ing temperature profile, the small temperature differences the temperature gradient dropped and the growth rate slowed, the that could arise among particles of various sizes are over- crystals rounded off and sintered, thus regaining some interparticle strength. whelmed by the overall gradient. Accordingly, the heat and vapor flows are predominantly upward in dry snow. These upward heat and vapor flows lead to the rapid disappearance of the small particles in dry snow, whereas in wet snow, the ture gradient and the growth rate and, accordingly, the small particles disappear because of heat flows among growth of faceted crystals ceases, producing both rounding neighboring grains of different temperatures. and sintering among the ice crystals.

4. Wet snow b. High liquid content Grain growth in wet snow above a critical liquid content (only Although most seasonal snow falls at a subfreezing tempera- about 7% by volume) occurs more quickly than at lower liquid ture, it is eventually wetted by rain and/or melt. Upon the contents. The melting temperature-particle size dependence introduction of the water, snow undergoes major changes in is stronger in the higher range of liquid contents and the liquid its grain morphology and in all of its important properties. area available for heat flow is greater. The transition between Like dry snow, wet snow can be described by two classes: well- these two liquid content categories, usually termed the pen- bonded and cohesionless. While dry snow undergoes a sharp dular (low) and funicular (high) regimes, occurs suddenly in transition at a critical growth rate (perceived as a critical any given pore. This sudden change in regimes is very impor- temperature gradient), wet snow undergoes a sharp transition tant in wet snow because it leads to equally sudden changes in at a critical liquid water content (Figs. 8 and 9). many important properties. The material strength is particu- larly important; it is reduced drastically at higher liquid con- tents. Grain clusters and their large grain boundaries are un- stable at high liquid contents; the clusters shown in Fig. 8 a. Low liquid content would break down into the cohesionless particles shown in Grain clusters form in wet snow with a low liquid content be- Fig. 9. Perhaps the most visible example of wet snow's lack of cause tightly bonded clusters minimize the total surface strength at large liquid contents is seen on highways. Vehicle energy in a material with all three types of interfaces: liquid- traffic over wet snow on road surfaces causes either rapid to-, vapor-to-solid, and liquid-to-vapor. Strong bonding compression of the snow into ice or splashing. At low liquid prevails within a cluster, although the clusters are not contents where wet snow develops considerable bond bonded as strongly to their neighbors. This allows wet snow strengths, compaction into ice is likely. However, at high liq- to retain a low density en masse, although the density of each uid contents where the snow is cohesionless, the slushy snow individual cluster may be close to that of solid ice. can be splashed aside easily. The formation of these clusters and the growth of grains in The transition to high liquid contents also is thought to be wet snow have long been attributed to melt-freeze cycles (Se- responsible for releasing wet snow avalanches. This probably ligman, 1936). This is unfortunate, since the clusters form occurs by water retention at a buried crust, followed by loss and the grains grow faster in the absence of temperature cy- of strength along that buried layer. cles. Clusters form and grains grow in order to minimize sur- The reason for the lack of grain boundaries at high liquid face free energy; they do so most quickly in the presence of contents apparently can be explained by the temperature de- liquid water and any amount of time spent below the melting pression caused by stress at the grain contacts. The contacts temperature can only slow matters down. As discussed in the act as stress risers, and crystalline boundaries are naturally

Unauthenticated | Downloaded 10/03/21 11:38 PM UTC 5 Vol. 64, No. 6, June 1983

FIG. 8. Metamorphic sequence starting with a wet snowflake. At a low liquid content, grain clusters develop rapidly in freely draining wet snow. The liquid is held along the grain boundaries and at liquid- filled veins at triple-grain junctions. The air temperature was above freezing throughout this period, a) Fresh snowflake (T — +2°C); b) 5 h old; c) 10 h old; d) 28 h old; e) 53 h old. weak anyway. The disintegration, or "candling," of lake ice along crystalline boundaries during the spring thaw and the cohesionless nature of ice particles in laboratory ice-water baths attest to this weakness. (If you are not familiar with these examples of cohesionless ice particles, examine the ice cubes in your next cocktail.) These observations, however, are different from Faraday's (1860) classic experimental ob- servation of suspended ice spheres. Faraday found that sus- pended spheres stick together even when placed in water. I once tried to explain this contradictory evidence on the basis of either impurities or electrical charges, but found that Fa- raday's conclusion was correct for all of the parameters I could control. Examine ice cubes submersed in water and try to reconcile your observation with Faraday's.

Unauthenticated | Downloaded 10/03/21 11:38 PM UTC Bulletin American Meteorological Society 607

FIG. 9. Cohesionless and well-rounded grains in wet snow at a high liquid content. Photographed through crossed polarized filters.

5. Mixed processes a. Melt-freeze grains

Although melt-freeze cycles are not responsible for grain growth or grain clusters in wet snow, these cycles do produce a distinctive type of grain in both seasonal and glacial snow covers. The individual particles in grain clusters like the ones shown in Figs. 8e and 10a are clearly distinguishable single crystals of ice. With repeated melt-freeze cycles, these crys- tals lose their individual character and visually become an amorphous version of the grain cluster (Fig. 10b). If subjected to strong solar radiation, these amorphous clusters can break down along their crystal boundaries and the individual crys- tals can reappear. FIG. 10. a) Grain cluster before any melt-freeze cycles. Each ice particle is a single crystal of ice. b) Amorphous form of grain cluster (or multicrystalline grain) produced by melt-freeze cycles. The indi- b. Surface melt crystals vidual crystals lose their visible identities.

Multicrystalline grains tend to break down into single crystals when subjected to solar radiation at the surface of a snow cover, which is why melting snow surfaces often consist of into smaller pieces and rapidly promotes rounding, much as cohesionless single crystals and tend to be weak and slushy. wave action does to beach sand. Unlike beach sand, however, In a strong solar radiation field, the surface layer of 1-2 cm wind crusts on snow are strengthened greatly by rapid sinter- can be scraped away easily, while the underlying layer, which ing—the deposition of vapor at the particle contacts, as consists of well-bonded grain clusters as shown in Figs. 8e shown in Fig. 11. These strong crusts are well known to skiers and 10a, is much stronger. This strong spatial gradient in who venture off the machine-packed trails. Of course, diurnal snow particle morphology and snow strength quickly disap- cycles of solar radiation can produce the same result, but by a very different mechanism. The production of very small pears along with the sun, a fact long used by skiers who rec- amounts of meltwater by solar radiation absorption in the top ognize the different snow surface conditions in the shadows few centimeters of the snow cover can cement the grains to- of trees. This surface disaggregation probably affects the al- gether quickly because all of the meltwater must flow to the bedo of the snow cover, since albedo generally is a strong bonds before refreezing. function of grain size, which varies greatly between slushy and well-bonded snows.

d. Melt-freeze layers c. Wind and sun crusts When surfaces are buried by subsequent snowfalls, they be- Wind-blown particles on the surface often form fine-grained come potential sites for meltwater retention and refreezing. crusts that subsequently are buried and give rise to the ice This happens most readily in shallow snow covers in temper- layers described later. The wind action breaks the particles ate climates, where the snow cover consists of many small

Unauthenticated | Downloaded 10/03/21 11:38 PM UTC 608 Vol. 64, No. 6, June 1983

water, at least during the onset of heavy melt in the spring. Like the melt-freeze grain cluster shown in Fig. 10b, ice layers break down under the influence of solar radiation. They also break down when the infiltrating meltwater is retained on their upper surfaces, mostly by enlargement of the triangular veins at triple-grain junctions, as shown in Fig. 12.

6. Summary

The snow falling from the atmosphere undergoes rapid changes as soon as it reaches the ground (Fig. 1). Snow parti- cles on the ground are as variable as, but very different from, snow crystals in the atmosphere. A wide variety of ice parti- cles from seasonal snow covers were shown here. These ice particles can be classified either according to shape or by the metamorphic processes that produce them. In Table 1 they FIG. 11. Wind crust crystals. These typically consist of small, have been classified according to the dominant process of well-rounded, and highly sintered particles. The arrow points to the formation or condition. crystalline boundary between these two ice crystals. Snow that has never been wetted takes on a well-rounded, equilibrium form at low growth rates and high temperatures (Fig. 2), and is well bonded by sintering. At large growth layers and melt periods typically are mixed with the refreezing rates (characterized by a large temperature gradient), snow of the entire snow deposit. There are several reasons for the recrystallizes, with highly faceted crystals (Fig. 3) replacing preferential retention of downward percolating water at the highly rounded ones. These highly faceted crystals are a these buried horizons. As explained previously, these hori- kinetic growth form of the ice crystal and lack interparticle zons are often wind crusts consisting of fine grains that hold strength because of reduced sintering. more capillary water in their small pores. Also, any difference Once wetted, snow takes on very different morphologies in pore size between layers can cause some liquid retention at and properties. This difference is most readily apparent in the base of the fine-textured layer. When subjected to refreez- the high-frequency dielectric behavior because of the large ing, this retained water further decreases the size of the pores dielectric constant of water as compared to that of ice or air. so that a positive feedback arises. The resulting "ice layers" There also are major differences within the category of wet are rarely impermeable, but do restrict the downward flow of snow, depending on the water content. At low water contents

FIG. 12. Etched crystalline boundaries on both the upper and lower surfaces of an ice layer in snow. A triangular liquid vein is shown at the junction of three grain boundaries (arrow). Trapped air bub- bles also are visible as large dark inclusions.

Unauthenticated | Downloaded 10/03/21 11:38 PM UTC Bulletin American Meteorological Society 609

(<7% by volume), wet snow is well bonded because the ice this shape is unstable. particles form into strongly bonded grain clusters (Figs. 8 Sintering—the movement of molecules to the interparticle and 10a). At high liquid contents, the snow tends to be cohe- contacts, which develop strength as they accumulate mass. sionless due to the weakening of the crystalline boundaries Most of the mass transfer is by vapor movement through (Fig. 9). the air. The higher vapor pressure over the rounded parti- Besides simply wet or dry snow, there are many processes cles is the driving force. that affect the morphology of the ice particles and the result- Snow crystal—an ice particle consisting of a single crystal of ing properties of the snow. Snow particles often are subjected ice. to a mixture of processes, which leads to a complicated grain Snow grain—an ice particle consisting of one or more crystals structure. Wind action, sunshine, melting, and freezing are of ice. the most important of these processes. For example, once Surface free energy—in the simplest sense, this is the energy grain clusters form in order to reduce surface free energy at required to create a unit area of additional surface area. low liquid contents, melt-freeze cycles can reduce these grain This usually varies with direction in a crystal. clusters to amorphous forms in which the original ice crystals Surface hoar—a highly faceted crystal that grows very rap- are no longer readily visible (Fig. 10b). Wind can break the idly on the surface during clear, cold nights. particles and promote rapid sintering, leading to strong Surface transition layer—a disordered surface or "liquid-like" crusts (Fig. 11). Melting and subsequent refreezing of buried layer that covers the crystalline surface in order to achieve crusts leads to semipermeable ice layers of increased density a stable transition between the ice and water vapor. The and strength (Fig. 12). thickness of this layer increases greatly as the melting A wide range of processes and conditions that lead to a temperature is approached. wide variety of ice particle morphologies and properties in Wet snow—snow at the melting temperature and containing seasonal snow covers have been shown. The emphasis here liquid water. The liquid water content is highly variable in has been on morphology, but much more information about wet snow. the properties is available (see the series of reviews published Wulff plot—given the surface free energy of every crystallo- with Colbeck, 1982). The processes that lead to these mor- graphic face, the shape of the crystal at equilibrium can be phologies and properties are being investigated and a deeper constructed using this simple technique. For an example, understanding is emerging. The different types of ice particles see Herring (1953). in snow were observed and classified on the basis of mor- phology some time ago. With increased knowledge of the underlying processes, it seems reasonable to classify them on the basis of causes rather than effects. No doubt this classifi- References cation will continue to evolve with our understanding of ice Bader, H., R. Haefeli, E. Bucher, J. Neher, O. Eckel, and C. Thams, particle morphology in the seasonal snow cover. 1939: Der Schnee und Siene Metamorphose. In Beitrage zur Geol- ogie der Schweiz, Geotechnische Serie, Hydrologie, Bern, Switzer- land. Bentley, W. A., and W. J. Humphreys, 1931: Snow Crystals. Acknowledgments. I thank Ed Wright and Dave Minsk for review- ing this manuscript. My work on snow has been supported by Proj- McGraw-Hill Book Co., New York. ect AT 161102AT24/C/EI/001 at CRREL. Colbeck, S. C., 1982: An overview of seasonal snow metamorphism. Rev. Geophys. Space Phys., 20, 45-61. de Quervain, M., 1954: Schnee als kristallines aggregat. Experimen- tal, I, 207-212. Appendix—Glossary of terms Faraday, M., 1860: Note on regelation. Proc. Roy. Soc. London, X, 440-450. Frank, F. C., 1982: Snow crystals. Contemp. Phys., 23, 3-22. Depth hoar—a highly faceted crystal that grows during periods Giddings, J. C., and E. LaChapelle, 1962: The formation rate of of high temperature gradients. These crystals develop depth hoar. J. Geophys. Res., 67, 2377-2383. most rapidly in the lower, warmer layers. Herring, C., 1953: The use of classical macroscopic concepts in sur- Equilibrium form—the shape that develops at equilibrium face-energy problems. In Structure and Properties of Solid Sur- when no crystal growth or shrinkage is occurring. This faces, edited by R. Gomer and C. S. Smith, University of Chicago term is used loosely here to describe crystals that grow Press, Chicago, pp. 5-72. slowly. Hobbs, P. V., 1974: Ice Physics. Clarendon Press, Oxford, England. Equilibrium thermodynamics—the basic assumption is that Kuzmin, P. P., 1957: Physical processes within a snow cover. In Fiz- icheskie Svoistva Snezhnogo Pokrova, Gidrometeoizdat, Lenin- the water, water vapor, and/or ice are in thermodynamic grad, pp. 11-26. equilibrium. This assumption is always approximate, since Nakaya, U., 1954: Snow Crystals. Harvard University Press, Cam- ice crystals in snow never stop growing or shrinking. bridge, Mass. Grain clusters—a tightly bonded group of snow grains Seligman, G., 1936: Snow Structure and Ski Fields. Macmillan and and/or crystals. Co., Ltd., London, England. Kinetic growth form—the shape that develops at high growth Wolley, J., 1858: Report of the British Association, Part II. pp. 40-41. rates when surface kinetics rather than thermodynamic Yosida, Z., et al., 1955: Physical studies of deposited snow. I. Ther- equilibrium controls shape. Once the rapid growth ceases mal properties. Low Temp. Sci., 7, 19-74. •

Unauthenticated | Downloaded 10/03/21 11:38 PM UTC